What is Lossy Compression?

Lossy Compression: Simple Definition

Lossy compression achieves small files by permanently discarding information. The decompressed result is not the original but an approximation, and the data removed is not recoverable by any means, because it no longer exists anywhere in the file.

In exchange the savings are dramatic. Where lossless compression of a photograph might halve it, lossy compression can reduce the same image to a small fraction of its size while looking essentially unchanged to a viewer, which is the entire reason the web is usable.

Quality settings are not a linear dial. Moving a JPEG from quality 95 to 85 often halves the file with almost no visible change; moving from 85 to 75 saves much less and costs much more. Most of the benefit sits in that first step, which is why blanket high-quality settings usually waste bandwidth.

Exploiting Human Perception

Lossy compression leans on perceptual coding: it models the limits of human sight and hearing and removes detail those senses are least likely to notice.^[1] JPEG transforms image blocks into frequency components and coarsely quantizes the high-frequency detail the eye perceives weakly, while audio coders discard sounds masked by louder nearby tones.^[2] The key idea is that "lossy" does not mean "low quality"; it means the loss is targeted at information you were never going to perceive, so a well-compressed lossy file can look or sound identical to the original at a fraction of the size.

Quality Settings and Artifacts

Most lossy encoders expose a quality or compression level that trades file size against fidelity. Pushing the setting too far introduces visible artifacts, blocking, ringing or color banding in images and a watery, hollow character in audio.^[2] The discarded information is gone permanently and cannot be restored by decompression.^[1] The skill is in choosing a setting just high enough that artifacts stay invisible, which for JPEG is typically around 80 percent quality and for audio around 192 to 256 kbps.

Generation Loss

Because each lossy pass throws away data, repeatedly editing and re-saving a lossy file degrades it cumulatively, a problem known as generation loss.^[2] Workflows therefore keep a lossless or original master and export to a lossy format only as a final step.^[1] This is why editing a JPEG ten times leaves it visibly worse than the original, while editing a lossless master and exporting once does not, the loss happens only at the moment of lossy encoding.

When Lossy Is the Right Choice

Lossy compression is the correct default for photographs, music, and video destined for viewing or listening rather than archiving or editing. For these, the file-size savings are enormous and the imperceptible loss is a fair trade. It is the wrong choice for text, code, data, and master files you will keep editing, where any loss corrupts meaning or compounds over time. The simple rule: lossy to deliver, lossless to preserve and edit.

How Lossy Compression Works

The trick is perceptual modelling. Rather than looking for patterns, a lossy encoder asks what a human will not notice. JPEG transforms the image into frequency components and discards the high-frequency detail the eye is least sensitive to. MP3 runs a psychoacoustic model and drops sound masked by louder sound nearby.

The consequence to understand is generational loss. Each re-encode runs the model over data that has already been stripped, so quality degrades with every save. Editing should happen in a lossless format, with a single lossy export at the end, and converting one lossy format to another always costs a generation for no gain.

The practical discipline is to keep a lossless master. Edit in PNG, TIFF, WAV, or FLAC, and export to a lossy format once, at the end, for delivery. Every conversion from one lossy format to another adds a generation of loss for no benefit, which is why MP3 to AAC is almost always a mistake.

Examples of Lossy Compression